FAQ#

from awpy import Demo

demo = Demo("match.dem")
print(demo.header)
print(demo.players)

Which games does Awpy support?#

Awpy reads Counter-Strike 2 demos in the PBDEMS2 format. It does not read CS:GO demos. For Deadlock, use Boon.

Which Python versions does Awpy support?#

Use Python 3.11–3.15. Awpy requires Polars 2.x. Wheels are available for Linux, macOS, and Windows. Installing a wheel does not require Rust. See Get started with Awpy.

How do I get Steam IDs and names?#

Use the resolved participants in demo.kills and demo.damages. Use demo.players for the roster. Raw demo.events fields such as attacker contain user IDs, not Steam IDs.

My demo has no round_start / round_end events — do rounds still work?#

Yes. Awpy reconstructs rounds from the CCSGameRules entity. This includes overtime rounds and does not require round events.

What are the winner numbers?#

2 means terrorist. 3 means counter-terrorist. Use winner_side for the name. Sides change at halftime; a side is not a team identity.

Why is demo.chat empty?#

Awpy can only return chat messages recorded in the demo. If it finds no supported messages, demo.chat returns an empty DataFrame with the chat schema. See Datasets.

When do map assets download, and can I work offline?#

Assets are cached under ~/.awpy. Set AWPY_DATA_DIR to change this path. Accessors use cached assets first and download missing assets. Set version= to select a release. Use awpy get or awpy.data.update() to check for updates. See Visibility & map data.

Why do datasets come back instantly the second time?#

Dataset properties and individual event DataFrames are cached on the Demo. Use demo.load(...) to collect compatible datasets in one pass. Unrelated datasets are not all prepared on first access.

player_stats() caches both values of include_knife_rounds. Its default result is also available as demo.stats. ticks() and snapshots() run for each call. Keep their results for reuse.

Create a new Demo to read the file without the existing caches.

How can I reduce parse time and memory use?#

  • Use event frames when you do not need entity state.

  • Use demo.load("kills", "damages", "shots") for related datasets.

  • Select only the properties needed in ticks(props).

  • Use snapshots() to limit output ticks.

  • Reuse cached DataFrames.

Entity decoding can still be required between selected ticks. A smaller result does not imply an equal reduction in parse time. See the benchmark guide.